1. Disease risk prediction
Usman Roshan

2. Disease risk prediction
Prediction of disease risk with genome wide association studies has yielded low accuracy for most diseases.
Family history competitive in most cases except for cancer (Do et. al., PLoS Genetics, 2012)

3. Disease risk prediction
Our own studies have shown limited accuracy with various machine learning methods
Univariate and multivariate feature selection
Multiple kernel learning
What accuracy can we achieve with machine learning methods applied to variants detected from whole exome data?

4. Chronic lymphocytic leukemia prediction with exome sequences and machine learning
We selected exome sequences of chronic lymphocytic leukemia from dbGaP. Largest at the time of download in August cases and 169 controls
Case and control prediction accuracy with genetic variants unknown
Same dataset previously studied in Wang et. al., NEJM, 2011 where new associated genes are reported but no risk prediction

5. What is whole exome data?
Human genome sequence
Introns
Coding regions
Exons
Illumina 76bp short reads (exome data).
In practice flanking regions are also sequenced
and so some intronic regions are included.

6. Obtain structural variants (1)
Human genome reference sequence
Data of size 3.2 Terrabytes and 140X coverage
Mapped to human genome reference with BWA MEM (popular short read mapper)
Short reads are aligned to human genome

7. Human genome
reference
Short reads from a
single individual
ATTAA
ACCAG
ACCAG
ATTGA
ATTGA
ACCAG
ACCAG
ATT--A
ATTGA
ACCAG
ACCAG
ATT--A
ATTGA
ACCCG
ACCAG
ATT--A
ATTGA
ACCCG
ACCAG
ATTGA
ATTGA
ACCAG
ATTGA
ATTGA
ATTGA
Here no variant is reported but we detected it in a different individual. Thus we assign it a value of 0 for this individual.
Heterozygous
SNP encoded
as 1
Homozygous SNP
encoded as 2 (0 if
same as reference)
Heterozygous indel
encoded as 1
Encoded into a feature
vector of four dimensions
(2, 1, 0, 1)

8. Obtain structural variants (2)
ATTGA
ACCAG
ATTGA
Obtained SNPs and indels from the alignments for each individual
ATTGA
ACCAG
ATT--A
ATTGA
ACCAG
ATT--A
ATTGA
ACCAG
Short reads from a
single individual
ATT--A
Human genome
reference
ATTGA
ACCCG
ATTGA
ATTGA
ACCCG
ATTGA
ATTGA
ACCCG
ATTGA
Homozygous SNP
encoded as 1 (0 if
same as reference)
Heterozygous SNP
encoded as 1
Heterozygous indel
encoded as 1

9. Obtain structural variants (3)
A/C C/G A/C C/G
C AA CC C
C AC CG C
C AA GG C
Co1 AC CG Co
Co2 CC CG Co
Combine variants from different individuals to form a data matrix
Each row is a case or control and each column is a variant
153 cases and 144 controls after excluding very large files and problematic datasets
SNPs and 2200 indels
Numerically encoded

10. Perform cross-validation study
Split rows randomly into training validation sets (90:10 ratio).
Rank all variants on training
Learn support vector machine classifer on training data with top k ranked variants
Predict case and control on validation data.
Compute error and repeat 100 times . . .
Training data
Validation data
Full dataset: each row
is a case or control
individual and each
column is a variant
(SNP or indel)

11. Variant ranking
F0 F1 F2 F1 F2 F0 C C C C C C Co Co Co Co
Rank features

12. Risk prediction with Pearson ranked SNPs
Similar curves with Pearson
Small k better than large k (k = number of variants)
SNPs better than indels
Top 60 SNPs mostly in chromosome 14

13. Prediction with GWAS

14. Cross-study validation

15. Prediction on external samples

16. Prediction on external samples

17. Pearson ranking of genes associated with CLL

18. Analysis of top ranked Pearson genes

19. Conclusion Encouraging results with exome data
No known risk prediction study with exome data
Limitations:
Small sample size
Ancestry of some data unknown